Mission Plan Recognition: Developing Smart Automated ...



Title: Cultural Agent Model to Predict inHabitant Opinion Reactions (CAMPHOR): Building and Applying a Dynamic Human Terrain Map

Suggested Topics: Modeling and Simulation

Authors:

Alexander Lubyansky

Aptima Inc.,

1726 M Street, N.W., Suite 900

Washington, DC 20036

Phone: (202) 842-1548x 338

Fax: (202) 842-2630

e-mail: alubyansky@

Yuri Levchuk

Aptima Inc.,

1726 M Street, N.W., Suite 900

Washington, DC 20036

Phone: (202) 842-1548x323

Fax: (202) 842-2630

e-mail: levchuk@

Georgiy M. Levchuk

Aptima Inc.,

12 Gill Street, Suite 1400

Woburn, MA 01801

Phone: 781-935-3966x267

Fax: 781-935-4385

e-mail: georgiy@

Krishna R. Pattipati

Professor, ECE Dept., UCONN

Storrs, CT

Phone: 860-486-2890

Fax: 860-486-5585

e-mail: krishna@engr.uconn.edu

Alexander Kott

DARPA / IXO

3701 Fairfax Drive

Arlington, VA 22203-1714

Phone: (571) 218-4649

e-mail: Alexander.Kott@darpa.mil

Correspondence:

Yuri Levchuk

Aptima Inc.,

1726 M Street, N.W., Suite 900

Washington, DC 20036

Phone: (202) 842-1548x323

Fax: (202) 842-2630

e-mail: levchuk@

Abstract

One of the strategic goals for the United States armed forces is to win over the hearts and minds of the population in a military theater of operations. Special Operations Forces (SOF) perform missions in part to help win this battle for hearts and minds. In order to be more effective in attaining this goal, SOF team leaders need easy and rapid access to accurate, timely, and detailed intelligence about the effect of mission actions on the opinions of the local population.

In this paper, we describe a methodology and a tool to improve the ability of intelligence officers to support SOF team leaders by providing an accurate and robust computer simulation of cultural dynamics that estimates the effect of SOF actions on local opinion. Social identity theory, theories from cognitive and social psychology, and theories related to social network analysis will inform the structure of the model. Data architecture will help users to populate the model even without complete, high-quality data. Experimentation and sensitivity testing will allow users to gain accurate insights when there is uncertainty in the data. With both model and data architecture, intelligence officers will be able to efficiently and effectively support SOF operations in pre-deployment, during mission planning, and in the field.

Motivation: Three Levels of SOF Planning Needs

“In accurately defining the contextual and cultural population of the battlespace, it became rapidly apparent that we needed to develop a keen understanding of the cultural intricacies that drive the Iraqi population.”

Major General (MG) Peter W. Chiarelli,

Commander, 1st Cavalry Division, Baghdad, 2004-2005

“…Most human behavior is learned observationally through modeling: from observing others one forms an idea of how new behaviors are performed, and on later occasions this coded information serves as a guide for action.”

–Albert Bandura, Social Learning Theory, 1977

One of the strategic goals for the United States armed forces is to win over the hearts and minds of the population in a military theater of operations. Special Operations Forces (SOF) perform missions in part to help win this battle for hearts and minds. The mission success in culturally-sensitive environments often hinges on the ability to acquire and apply robust knowledge of socio-cultural phenomena that underlies the human terrain environment of the battlespace – often very complex and difficult to navigate. The human terrain presents challenges of compressed tempo, nonlinear reactions to culturally loaded events (e.g., religious beliefs, customs, and norms), and interactions across complex networks of family, tribe, party, economy, and nation. For instance, SOF must consider how both combat and non-kinetic actions may interact with other influences (such as the media, non-government organizations, and basic supplies such as food) to predict the effects of specific COA that render the mission success or failure.

Tools such as Pentagon's Human Terrain System (HTS) -- which involves social scientists helping the military understand local culture and politics -- are being designed to better equip the military with cultural knowledge. In its current conception, HTS is built upon seven components, or "pillars": human terrain teams (HTTs), reachback research cells, subject-matter expert networks, a tool kit, techniques, human terrain information, and specialized training. The core building block of the HTS is a five-person Human Terrain Team (HTT) that is embedded in each forward-deployed brigade or regimental staff. The HTT provides the commander with experienced officers, NCOs, and civilian social scientists trained and skilled in cultural data research and analysis. However, despite recent technological advances in simulations and modeling capabilities, tools that are needed to support the HTS fall short in several respects: (a) Theory – existing tools lack a theoretical foundation for categorizing and describing culturally important behaviors, necessary to make cultural training systems useful across sub-cultures (e.g., Sunni and Shiite) and cultures (e.g., Iraqi and Afhani); (b) Data – users lack tools and methods for capturing cultural behaviors; and (c) Interfaces – the displays and controls do not support crucial behaviors (cues and responses) and contexts that give them meaning.

When planning for counter-insurgency operations in culturally-sensitive environments, the SOF faces the three related problems:

1. The organizational problem of improving organizational knowledge and institutional memory related to counterinsurgency and associated environment (e.g., acquiring knowledge of socio-cultural context and transferring such knowledge to the newly arrived forces).

2. The tactical problem of planning a counterinsurgency mission in a given area (e.g., such as a village).

3. The strategic problem of conducting multiple counterinsurgency missions in a broader theater (e.g., that comprises several villages).

In order to accomplish planning, the SOF team leader needs rapid access to understandable, accurate, and usable intelligence, including intelligence about how actions taken during the mission may affect the sentiments of the local population. The complexity of the cultural dynamics of the local population makes it difficult for intelligence officers to conduct deep and rigorous analysis of the cultural effects of SOF team actions in the field.

Solution: Cultural Agent Model to Predict inHabitant Opinion Reactions

“Culture is the ‘human terrain’ of warfare. Human terrain is the key terrain.”

MG(Ret) Geoffrey Lambert

“…Navigate cultural and human terrain as easily as

Marines can now use a map to navigate physical terrain.”

LtGen James Mattis

To address the above challenges, we have developed a prototype of the Cultural Agent Model to Predict inHabitant Opinion Reactions (CAMPHOR) system that will:

a) Encode relevant human terrain factors that can be the key to successful mission planning and execution;

b) Store the acquired socio-cultural knowledge and present it to the human users in a concise and intuitive manner as understandable, accurate, and usable intelligence;

c) Allow for rapid analysis and prediction of probable effects of culturally-sensitive courses of actions (COA) that may affect the sentiments of the local population and promote or inhibit the mission success; and

d) Help select the steps necessary to deal appropriately with the insurgencies within the context of their unique cultural environments.

The CAMPHOR will help the SOF mission planners to tackle the important cultural and social dynamics that are typically too complex for the unaided human mind to analyze accurately. The ultimate goal of the CAMPHOR framework is the design of advanced collaborative planning tools to predict the range of possible RED and GREEN behaviors within the theater of operations, in order to support the development of the SOF COA for counter-insurgency operations in culturally-sensitive environments. CAMPHOR will also enable the pre-deployment mission training of rotating SOF and other US military forces. CAMPHOR will fill the cultural knowledge void by gathering ethnographic, economic, and cultural data pertaining to the battlefield to support analysis and decision-making and to provide the SOF commanders with a fully automated comprehensive cultural information research system.

Solution Overview

The CAMPHOR system offers the following functionality (Figure 1):

1. Knowledge Repository, in the form of Human Terrain Maps, for encoding and archiving of the socio-cultural artifacts to support counter-insurgency planning in specific localities and support a broader use of socio-cultural Intelligence within the theater of operations.

2. Data Fusion Algorithms for generating the Human Terrain Map (HTM) of a given locality (e.g., a village) from the available Intelligence data, including algorithms for automatically generating the many elements of the HTM (e.g., automatically reverse-engineering power and communication networks from ISR observations and intercepted communications and transaction data).

3. An executable dynamic “cultural map” of the mission environment (i.e., a computer simulation component of the Human Terrain Map) that allows one to predict how various SOF COA may interact with the local socio-cultural environment (e.g., produce changes in local population opinions), as well as predict the concomitant range of possible RED and GREEN behaviors and probable scenarios that may result from specific COA.

4. User Interface for navigating the Human Terrain Map to help intelligence officers rapidly input data; conduct analyses; understand outputs; and communicate results to the SOF team leaders. During the pre-deployment mission training/rehearsal, the User Interface will facilitate compact communication of the critical socio-cultural background information and of the lessons learned.

5. Algorithms for inferring and analyzing the high payoff COA options and for developing efficient counter-insurgency mission plans.

[pic]

Figure 1. CAMPHOR solution workflow.

In a nutshell, CAMPHOR transforms the culturally relevant bits and pieces of information from various sources (e.g., from news feeds, historical archive data, satellite intercepts of electronic communications, and Intel from SOF deployed in theater) into a comprehensive Human Terrain Map that encodes the relevant socio-cultural knowledge and provides automated means for predicting cultural and social dynamics that are typically too complex for the unaided human mind to analyze accurately (Figure 2). This information is than used to infer the mission threats and opportunities and to appropriately optimize the selection of COA, in order to achieve superior performance in counter-insurgency operations and “hearts and minds” campaigns.

[pic]

Figure 2. CAMPHOR – the “Big Picture”.

Figure 3 outlines how CAMPHOR builds up a navigation-capable Human Terrain Map. The automated data fusion methods and algorithms – to be combined with the manual data entry, if desired – that enable steps depicted in Figure 3, are discussed in the following (sub)section (‘Data Sources and Fusion Methods’; also see more details in ‘Human Terrain Map’).

[pic]

Figure 3. How CAMPHOR constructs navigation-ready Human Terrain Map – an overview.

Figure 4 depicts the CAMPHOR Concept of Operations that outlines the CAMPHOR system concept (inputs / outputs; hands-on users; and measures of the CAMPHOR process effectiveness and efficiency). It shows that the inputs from various Intel and data sources are combined to generate/refine the Human Terrain Map, which is than used by the Intel analyst (who assists the mission planning) to predict threats and opportunities and to select COA that optimize Measures of Effectiveness/Performance.

[pic]

Figure 4. CAMPHOR Concept of Operations.

The CAMPHOR inputs include: (1) Assumptions about the local environment and concomitant material conditions; (2) Available culturally-relevant up-to-date (static) intelligence; (3) Mission planning needs (objectives, goals, success criteria); (4) Mission candidate actions; (5) Contextual Intel – recent events in the village. The CAMPHOR outputs include: (A) Human Terrain Map (groups, individuals, networks, their key socio-cultural attributes and behavior envelops); (B) Probable events (including threats and opportunities) for each given COA; and (C) Mission Plans (e.g., OPS Order) composed from high-value COA selections.

Data Sources and Fusion Methods

“The information we desire the most about the enemy-his real fighting power and his intentions-lie in the psychological and human dimensions rather than the physical...”

-- Dr. Ahmed S. Hashim

U.S. Naval War College

One of the important steps in the doctrinal military decision-making process (MDMP) (Wade, 2005) through which the U.S. Army conducts its operations is the intelligence preparation of the battlefield (IPB), The IPB requires the assessment of enemy’s command and control structure to predict the actions of the adversary, identify high-value targets, and develop effective counteractions. Currently, the intelligence operations officer provides input to help the planning officer develop the IPB templates, databases, and other products that portray information about the adversary and other key groups in the area of operations or area of interest (Figure 5). The IPB products contain information about each group’s leaders and decision makers, size and location of enemy forces, and linkages among groups and leaders. The linkage information is produced manually from the data on activities (using activity matrix template) and intelligence on the relationships between individuals. Using this information, a link diagram is developed to show the interrelationships of individuals, organizations, and activities. The concomitant information from the IPB products about the key groups and their leaders (and about linkages among individuals) will serve as the good initial starting point to provide the input for the CAMPHOR system.

Figure 5. Sample Information Operations doctrinal template (adapted from (FM 3-13, 2003)).

Currently, only a limited set of tools is available to intelligence operators to analyze, correlate and visualize the data. The two most commonly used network analysis tools are StarLight () and AnalystNotebook ( Analysts_Notebook/default.asp). These tools are often used together with technologies performing data mining and automated entity and link discovery from text sources (Miller et al., 2000; Grishman, 2003; Stolfo et al., 2003) or manually using HUMINT and other data sources. They rely on domain understanding (Krebs, 2001; Sageman, 2004) or applied social network analyses (Van Meeter, 2001; Dombroski and Carley, 2002; Dombroski, Fischbeck, and Carley, 2003; Skillicorn, 2004). However, these tools merely present and visualize the networks formed by observations and do not solve the network identification problem of “cleaning uncertain observations”, especially in the presence of missing, irrelevant, deceptive, and mislabeled attributes and links. As a result, current approaches to analyzing networks are manual: the analysts rely on their experience to make sense of visualized structural and temporal data. Large information gaps, including missing data, deceptions, and errors, have to be dealt with, and analysts often fill the gaps with their experience, which may not be applicable to the problem they need to solve, thereby resulting in decision biases. In addition, people tend to exhibit confirmatory biases when the first seemingly valid hypothesis is selected and further relied upon during the analysis. This issue is compounded by huge amounts of data and complexity of the problem people need to analyze, influencing what data is used and which is filtered out and thus never studied. All these factors negatively impact the ability of the intelligence team to recognize a potential enemy or other key groups and further result in decreased effectiveness of counteractions and unintended consequences.

CAMPHOR, however, will seek to automate the reverse engineering of linkage information (see Figure 6 for the problem outline) and other essential socio-cultural factors for the key groups in a given area of operations locality, based on the data feeds from various information sources (e.g., from phase zero, open-source, baseline socio-cultural information to focused ISR observations of select individuals and organizations). Although human actions leave in the information space detectable events whose dynamic evolution creates patterns of the potential realization of activities that may be related, linked, and tracked over time (Pattipati et al., 2004), the socio-cultural data is often very sparse, creating a challenge to connect relatively few relevant events embedded within massive amounts of data. To extract critical information from the ISR data, CAMPHOR customizes algorithms originally developed by APTIMA and successfully validated under the DARPA NetSTAR (Identification of Network, Structure, Tasks, Activities, & Roles from observations) project (contract HR0011-06-C-0047, DARPA, 1/27/06 – 7/27/06; DARPA TPOC Alex Kott: Alexander.Kott@darpa.mil).

[pic]

Figure 6. Network identification problem setup.

NetSTAR uses variants of Hidden Markov Models and Hidden Markov Random Fields, to reverse-engineer communication and other networks that link individuals from multi-source uncertain ISR data based on probabilistic attributed network pattern matching principles (Levchuk et al., 2007). From ISR observations about individual actions, acting entities and concomitant interactions and transactions, NetSTAR reverse-engineers the missions (goals and tasks), relationships, roles of individuals, their functions, tasks performed, and categories/ frequencies of interactions (e.g., information exchanges, command relationships, meetings, financial transactions, etc.). Experiments have shown that NetSTAR provides reliable state prediction of the organizational network under high levels of missing data, sensor errors, and deceptive or irrelevant observations (Figure 7; see also Levchuk et al., 2006 and 2007). NetSTAR is capable of inferring the principles and goals under which the groups of interest operate and thus can facilitate predictions of the group’s intent and likely behaviors.

[pic]

Figure 7. NetSTAR algorithm validation results.

Equipped with NetSTAR algorithms, CAMPHOR will aim for seamless integration of data from disparate sources (e.g., to combine two or more sources of complementary data). Ultimately, in addition to the dynamic on-the-ground Intel and satellite intercepts feeds, the CAMPHOR system will draw on data streams from international and domestic sources; press and news services; interviews; press releases; international peacekeeping, lending, and aid agencies; analyses from regional and technical experts, and other sources. Finally, in Phase II we will examine how CAMPHOR may interact with centralized socio-cultural knowledge databases, such as those presently developed under the Pentagon's Human Terrain System initiative.

Human Terrain Map

“Cultural awareness will not necessarily always enable us to predict what the enemy and noncombatants will do, but it will help us better understand what motivates them, what is important to the host nation in which we serve, and how we can either elicit the support of the population or at least diminish their support and aid to the enemy.”

Major General Benjamin C. Freakley,

Commanding General, CJTF-76, Afghanistan, 2006

The key component of the CAMPHOR system is the Human Terrain Map (HTM), which formally represents and organizes the relevant, complex socio-cultural knowledge of the psychological and human dimensions of the theater of operations. As will be shown below, the HTM links the socio-cultural knowledge about the area of operations to the physical landscape and to geographical and geopolitical infrastructure.

The concept for the “human terrain” (McFate and Jackson, 2005) defines the socio-cultural, anthropologic, and ethnographic data and other non-geophysical information – geospatially and relationally referenced (and, in some cases, temporally referenced) – about human population and society in the area of operations. It includes the situational roles, goals, relationships, and rules of behavior of operationally relevant groups and/or individuals. It also includes the underlying formal and informal structures (e.g., relatives and friendship circles, communication and social networks) that shape interactions, knowledge and information transfer, and behaviors of groups and/or individuals.

In a strict mathematical sense, the CAMPHOR Human Terrain Map is a multilayered multi-attribute graph (Figure 8) that encodes the relationships and interactions between groups, between key individuals, and between individuals and groups, while also linking the individuals and groups to facilities with fixed geo-locations (Figure 9). Hence, the HTP links the individuals and groups to geographic areas (encompassing all the facilities that a given individual or a group is associated with). The HTP navigation exploits the reverse mapping between geography and groups/individuals. Because of potentially different strength of associations between an individual (and/or a group) and various facilities, geographic areas may in turn have different strength of associations with different individuals (and/or groups) – analogous to how geographic areas have different terrain elevation. The attributes of the HTP represent the group-specific cultures -- relatively stable but with the potential to change over time -- the top-most HTP graph attributes represent the culture shared across a society. Also, the attributes associated with individuals (i.e., the concomitant node attributes) encode individual differences; the degree to which individuals adopt and practice the culturally shared attitudes, values, beliefs, and behavior tendencies; and their unique goals, aspirations, and significant behavior tendencies (if any).

[pic]

Figure 8. CAMPHOR “under-the-hood” HTM construction process – an overview.

[pic]

Figure 9. Extracting facilities and linking them to individuals, groups – an overview.

The CAMPHOR HTM framework: (1) helps categorize and profile the local population and capture culturally important behaviors and motivations, cultural cues and other behavior shaping factors, and the contexts in which they have meaning; (2) focuses the concomitant Intelligence gathering; (3) helps fuse the socio-cultural information and links it together into a comprehensive knowledge base; (4) guides and automates the analysis of how various SOF COA may impact the opinions of the population, and promote or inhibit specific behaviors; (5) automates the process of predicting the concomitant range of possible RED and GREEN behaviors and probable scenarios that may result from specific COA; (6) helps communicate to the SOF team leaders and other decision-makers the critical socio-cultural background information and the lessons learned; and (7) facilitates pre-deployment mission training and rehearsal by the SOF teams arriving in theater.

The users will be getting the information from CAMPHOR in the form of IPB-like documents, namely -- the socio-cultural terrain briefs / SCTB (Figure 10; also see the detailed description of an SCTB below). The users will receive a focused SCTB about any geographic area they choose (from within the area for which the HTP has been constructed and populated). The socio-cultural terrain brief will provide the intuitive interface between CAMPHOR and its users, as it will present in a concise and clear way and in a free-flow text format the information stored within the multilayered multi-attribute HTP graph. The “under-the-hood” structure of HTM that will help the users navigate across the HTM, will be largely invisible to the users (but it will inform the subsequent automated analysis of the COA effects).

[pic]

Figure 10. CAMPHOR socio-cultural terrain brief – an overview.

The CAMPHOR Human Terrain Map, and its socio-cultural terrain brief (outlined in Figure 10; see also the detailed illustrations of SCTB in Use Case examples described below), encodes the following information about human population in the area of operations:

1. Social and material conditions. This section of the HTM provides the basic demographics data (e.g., the approximate number of families or households in a village); describes the key parts of the villagers’ economy; and lists other relevant material conditions under which social arrangements develop.

2. Group composition and profiles. This section lists the key identified influential groups, and profiles the group-specific culture artifacts (e.g., values, norms, objectives, beliefs) that shape the group behavior (later in this section, we describe the algorithms for predicting the range of group’s behaviors). At the societal level, culture artifacts involve mainstream average tendencies, while at the individual level they represent the level of individual’s participation in shared values, beliefs, and behavior tendencies. The balance between the within-group culture variance and the between-group culture variance is an indicator of stability within a given society (or group). The cultural artifacts help infer the competing agendas, balance of power, opportunities exploitable from group’s viewpoint, and risks of cultural confrontation that pit alternative value systems against each other.

– Beliefs, values, and norms. For a given group, this subsection lists the key known culturally shared traditions; assumptions of what constitutes “good” vs. “bad” behavior, “desirable” vs. “undesirable” practices, “fair” vs. “unfair” actions in a given situation; expectations about other groups; and attitudes in dealing with people from the own group and from other groups.

– Objectives. For a given group, this subsection of the HTM lists the key known commonly-shared objectives and needs of group participants, including the explicit goals (if any) expressed by the group leaders.

– Stereotypical behaviors. For a given group, this subsection lists the common “trade-mark” individual and collective actions of group members in specific situations. Primarily, this involves repeatable behavior patterns that are likely to occur in the future.

3. Individual profiles. This section lists people of interest in a village, and profiles their group identities (when known) and individual-level cultural artifacts (e.g., values, goals, aspirations), emphasizing individual differences with the culture of the groups they belong to or identify with.

– Group identities. For a given person, this subsection lists the known groups that this individual belongs to or identifies with (there could be several such groups).

– Aspirations and goals. For a given person, this subsection lists his/her known individual goals (e.g., when he or she is studying to become a doctor; is generally looking for opportunities to make a living; desires to send his/her children to school; is in need of a specific medical treatment; etc.).

– Personal characteristics. For a given person, this subsection of the HTM lists factors that shape (and may be exploited to influence) the behavior of a given individual.

– “Trade-mark” behavior traits. For a given person, this subsection lists his/her unique personal behavior traits (if any); it may also list the known repeatable (fixed) action patterns and historical events that involved a given individual.

– Cognitive characteristics. For a given person, when known, this subsection indicates the education level and other factors (e.g., risk-aversiveness) that can be used to predict decision-making trends and capabilities.

4. Networks. This section of the HTM displays the social connections (e.g., work relationships, frequent face-to-face communications, propinquity/friendship and kinship relations, supply chains, etc.) that link people in a village, as well as connections between various groups and between individuals and groups. While the data about networks stored in CAMPHOR can in most cases be incomplete/partial (especially given the dynamic nature of many networks), it nevertheless provides the useful information about the network-driven social organization of the society and/or some of its segments (e.g., it can show pathways for disseminating information; who may benefit from being well connected; whether the critical mass of interactions is reached; and so on). The concomitant network structures are used by CAMPHOR: (a) to determine the key individuals (e.g., the most influential individuals or those who control the information; people with a large number of interpersonal interactions with others, or those involved into the inter-group ties); (b) to analyze the information flow and collaboration boundaries; and (c) to detect the emergent groups.

– Power structure. This section of the HTM encodes in a graph format the subordination relationships (e.g., superior-subordinate work relationships; teacher-follower relationships; etc.) that link individuals. It also implicitly encodes the pressure points and channels of influence that can be exploited to shape the behavior of groups and individuals.

– Ego networks. This section encodes in a graph format the circle of people that a given person interacts with frequently (e.g., near-daily).

– Kinship and friendship circles. This section encodes in a graph format the circle of people that a given person is closely-related to (i.e., close relatives) or is emotionally attached to (i.e., close friends). The types of links between people distinguish the corresponding relationships (kinship vs. friendship); also, the strength of the relationship (e.g., kinship closeness; frequency of interactions with a friend) may be encoded as a separate link attribute.

– “Etc.”: Other types of relationships may be encoded in CAMPHOR in the future.

Figures 11 and 12 illustrate the elements of a Socio-cultural Terrain Brief (see Use Cases below for more details).

[pic]

Figure 11. Elements of a socio-cultural terrain brief – material conditions and group profiles – an illustration.

[pic]

Figure 12. Elements of a socio-cultural terrain brief – facilities, networks, and individual profiles – an illustration.

Predicting behaviors

The HTM and its culture-specific attributes -- from objectives to resources to social structure to stereotypical behaviors (prototypical action patterns) -- facilitate predictions of behavior trends (e.g., for preferring/choosing specific actions) of organizations, groups, and individuals in response to a specific stimuli (e.g., COA) or situation. Figure 13 outlines the method by which these predictions are carried out (based on the “difference engine” and “general problem solver” principles combined with bounded rationality principles; e.g., see Minsky 1985; Kahnemann 2002; Smith 2002).

The group objectives and/or individual goals define the “desired” situation, various aspects of which (e.g., subgoals) “arise” different elements of the social system (e.g., individuals, groups, information pathways, social pressure) that then act (i.e., select and perform actions) to change the state of the world they control or influence in a way that tends to diminish the difference that aroused it. In essence, those behaviors and interactions that benefit the objectives/goals tend to be preserved and/or reinforced, while other behaviors and interactions that adversely affect objectives/goals tend to be avoided and/or reduced. This tendency is mediated by bounded rationality (e.g., “satisficing” rather than optimizing; oscillating between reactive and proactive decision-making; shifting between risk-taking and risk-avoidance based on the goal state and situation) that affects probabilistic choices.

Remarkably, the corresponding behaviors by various groups and individuals need not be conscious or directed; very often, such behaviors are “self-organizing” and emerge from the interactions among the many “agencies” that become engaged in pursuit of the concomitant (sub)goals (Smith, 2002, Nobel Price lecture). As will be shown in the Use Case examples below, these principles may be systematically used during the SOF mission planning to discover/select COA that would promote objectives/goals of certain groups and stimulate behaviors that promote the success of the SOF mission.

[pic]

Figure 13. CAMPHOR predictions of behavior trends – an overview.

The CAMPHOR will help the SOF mission planners in predicting the cultural and social dynamics that are too complex for unaided human mind to analyze accurately -- especially when dealing with non-commensurate and counteracting objective systems and social pressures that interact in a complex, indirect fashion. Goal-driven systems often do not react directly to the stimuli or situations they encounter. Instead, they treat the things they deal with as objects to explore, avoid, or ignore, as though they are concerned with something that doesn’t yet exist (e.g., with subgoals, “ideal” inputs, favorable constraints, etc.). When any disturbance or obstacle diverts a goal-directed system from its course, that system seemingly tries to remove the interference, go around it, or turn it to some advantage.

While the objectives/goals indicate the predisposition of a person or group to choose from amongst its prototypical actions (or to acquire/learn/try new actions), networks provide flexible means of social organization that constrains and shapes behaviour (e.g., by constraining access to information and mobilizable social capital; mediating supply chains and options for rapid dissemination of outputs; fostering human interactions and material or emotional support; etc.). Also, agent’s resources will impact the attractiveness of each action type (e.g., based on the cost-benefit analysis) that will impact the action selection in Figure 13. The specific network structure can mediate the cost/benefit of rapidly acquiring resources, and thus impact the CAMPHOR behavior predictions. Networks exhibit emergent properties of structure and composition by “rewarding” certain behaviors that are compatible with their ecology (and by “sanctioning” behaviors that are mismatched with the ecology of the concomitant cultures). Figure 14 illustrates the objectives/goals and actions inputs into CAMPHOR for a hypothetical Iraqi Insurgency Scenario.

[pic]

Figure 14. Iraqi Insurgency Scenario goals and actions inputs into CAMPHOR (derived from Intel and historical data) – a hypothetical example.

Predicting effects and scenarios

CAMPHOR will use its HTM (with its underlying structures and culture-specific attributes) -- together with the candidate COA by the SOF and US forces -- to construct a dynamic executable model of the mission environment and its socio-cultural terrain. This model is a hybrid model that includes agent-based, system dynamics, and social network components. Figure 15 illustrates the process of building a mixed system dynamics / agent model in CAMPHOR from a socio-cultural terrain brief, while analyzing options for COA directed at different groups (see Use Cases below for more details regarding this example).

[pic]

Figure 15. Building a mixed system dynamics / agent model in CAMPHOR from a socio-cultural terrain brief – an illustration.

The CAMPHOR dynamic model of the socio-cultural terrain enables Intelligence Analysts to conduct on-the-fly assessment of COA effectiveness for the mission at hand (the richness of available data determines the uncertainty and the breadth and focus of the predictions envelop for the COA testing). The CAMPHOR dynamic model assesses possible effects of the COA on the opinions of the local population and on the likelihood of attitude changes by different groups and key individuals.

More specifically, the CAMPHOR dynamic model addresses the following facets of human individual and group behavior:

1. Cognitive processes of people who perceive events, judge their meaning, memorize past events, and (possibly) learn from their experiences.

2. Cultural processes of people who possess and change their opinions about other people, concomitant relationships, and affiliations with different identity groups. For example, dynamic shifts in allegiances to specific social and cultural identities could manifest the behavior specific potential of terrorist groups to seek affiliation with the jihad, and could indicate their intent to commit atrocities against western nations (Sageman, 2004).

3. Communication processes people use to spread information about events. Theories of social influence (Asch, 1955; Festinger, 1954; Friedkin, 1999; Milgram, 1974) suggest that the network structure in which communications take place is very important in determining how communication changes people’s opinions. Social network analysis (Wasserman & Faust, 1999) that studies the effect of networks on human behavior, identifies the frequency of communication as the key observable variable that determines the process of social influence.

4. Event characterization (e.g., social event characterization scheme) that is related to classes of observable activities and artifacts that are encountered in the physical world (e.g., an event such as marriage exposes people to new kinship and friendship networks, which may inspire their affiliation with the jihad).

The CAMPHOR dynamic model facilitates predictions of probable event scenarios driven by applying specific time-stamped COA (as illustrated in Figure 16). When the assessed propensity for a specific event exceeds the corresponding threshold, the model then generates an event and records its instance, to produce the scenario that is fed dynamically back into the model to assess the concomitant propagation of effects (e.g., attitude, opinion, and/or allegiance changes; abandoning goals that have been achieved and reprioritizing objectives; engaging in reactive behavior; etc.). While the Monte-Carlo runs are needed for generating comprehensive predictions envelop, even a single CAMPHOR dynamic model run produces a probable scenario that can be used to adjust the mission planning in real time.

[pic]

Figure 16. CAMPHOR predictions of scenario events – an illustration.

Intelligence Analysts will compactly communicate the lessons learned to the SOF team leader both during mission planning, and when the SOF team is deployed in the field. The SOF team leader will be able to get periodic updates about the probable effects of the relevant mission developments on the opinions of the local population.

References:

(Wade, 2005) Wade, N.M., The Battle Staff SMARTbook: Doctrinal Guide to Military Decision Making and Tactical Operations, 2nd Edition, Lakeland: The Lightning Press, 2005

(Miller et al., 2000) Miller, D., S. Boisen, R. Schwartz, R. Stone, R. Weischedel, “Named entity extraction from noisy input: speech and OCR”, Proceedings of the sixth conference on Applied natural language processing, Seattle, Washington, 2000, pp. 316-324

(Grishman, 2003) Grishman, R., “Information extraction.” In R. Mitov (ed.), The Oxford handbook of computational linguistics, pp. 545-759, New York, NY: Oxford University Press, 2003

(Stolfo et al., 2003) Stolfo, S.J., et al., “Behavior Profiling of Email,” Proceedings of NSF/NIJ Symposium on Intelligence & Security Informatics, 2003

(Krebs, 2001) Krebs, V.E., “Mapping Networks of Terrorist Cells”, Connections, Vol. 24, 2001, pp. 43–52

(Sageman, 2004) Sageman, M., Understanding Terror Networks, Philadelphia, PA: University of Pennsylvania Press, 2004

(Van Meeter, 2001) Van Meeter, K.M., “Terrorists/Liberators: Researching and Dealing with Adversary Social Networks”, Connections, Vol. 24, 2001, pp. 66-78

(Dombroski and Carley, 2002) Dombroski, M.J., and K.M. Carley, “NETEST: Estimating a Terrorist Network's Structure”, CASOS 2002 Conference, No. 8, 2002, pp. 235-241

(Dombroski, Fischbeck, and Carley, 2003) Dombroski, M.J., P. Fischbeck, and K.M. Carley, “Estimating the Shape of Covert Networks”, Proceedings of the 8th International Command and Control Research and Technology Symposium, National Defense War College, Washington, DC, 2003

(Skillicorn, 2004) Skillicorn, D., “Social Network Analyses via Matrix Decompositions: al Qaeda”, Report, available from , Aug., 2004

(Pattipati et al., 2004) Pattipati, K., Willett, P., Allanach, J., Tu, H., and Singh, S. “Hidden Markov Models and Bayesian Networks for Counter-Terrorism”, in “21st Century Enabling Technologies and Policies for Counter-Terrorism”, Robert Popp and John Yen eds. (2004).

(Levchuk et al., 2007) Levchuk., G., F. Yu, H. Tu, K. Pattipati, Y. Levchuk, and E. Entin, “Identifying the Enemy – Part I: Automated Network Identification Model”, to appear in Proceedings of the Command and Control Research and Technology Symposium, New Port, RI, 2007.

Levchuk, G., Yu, F., Meirina, C., Singh, S., Levchuk, Y., Pattipati, K., Willett, P., and Kelton, K. “Learning from the Enemy: Approaches to Identifying and Modeling the Hidden Enemy Organization”, in A. Kott (ed), Information Warfare and Organizational Decision-making, Artech House, Norwood, MA, 2007

(Levchuk et al., 2006) G. Levchuk, Y. Levchuk, and K. Pattipati, “Identifying Command, Control and Communication Networks from Interactions and Activities Observations”, Command and Control Research and Technology Symposium, 2006, San Diego, CA.

(McFate and Jackson, 2005) McFate and Jackson, An Organizational Solution for DOD’s Cultural Knowledge Needs, MILITARY REVIEW, July-August 2005 ()

(Minsky 1985) M. Minsky. Society of Mind Journal of Semantics 1986 5(2):163; doi:10.1093/jos/5.2.163, 1986 by Oxford University Press

(Kahnemann 2002) Daniel Kahnemann, MAPS OF BOUNDED RATIONALITY: A PERSPECTIVE ON INTUITIVE JUDGMENT AND CHOICE, Nobel Prize Lecture, December 8, 2002

(Smith 2002) Vernon Smith. CONSTRUCTIVIST AND ECOLOGICAL RATIONALITY IN ECONOMICS, Nobel Prize Lecture, December 8, 2002

(Asch, 1955) Asch, S. E. (1955). Opinions and Social Pressure. Scientific American, pp. 31-35.

(Festinger, 1954) Festinger, L. (1954) A Theory of Social Comparison Processes. Human Relations 7, 117-40.

(Friedkin, 1999) Friedkin, N. E. (1999). Choice Shift and Group Polarization. American Sociological Review, 64(6), 856-875.

(Milgram, 1974) Milgram, S. (1974). Obedience to Authority. New York: Harper and Row.

(Wasserman & Faust, 1999) Wasserman, S. & Faust, K. (1999). Social Network Analysis: Methods and Applications. Cambridge: Cambridge University Press.

Levchuk, G.M., K. Chopra (2005), “NetSTAR: Identification of Network Structure, Tasks, Activities, and Roles from Communications”, Proceedings of the 10th International Command and Control Research and Technology Symposium, McLean, VA, June.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download